Task for predicting ICU mortality using eICU-CRD dataset. Predicts whether a patient will die during the ICU stay based on clinical information collected during the stay.
Details
This task uses the unit discharge status from the patient table to determine mortality. The prediction is based on clinical codes from diagnosis, physicalExam (procedures), and medication tables.
Label definition:
Label = 1 if unitdischargestatus == "Expired"
Label = 0 otherwise
The task excludes:
ICU stays without any clinical codes (conditions, procedures, or drugs)
Features:
conditions: ICD-9 codes from diagnosis table
procedures: Physical examination paths from physicalexam table
drugs: Drug names from medication table
Super class
RHealth::BaseTask -> MortalityPredictionEICU
Methods
Method call()
Process a single patient (ICU stay) to generate samples.
Examples
if (FALSE) { # \dontrun{
library(RHealth)
# Load eICU dataset
eicu_ds <- eICUDataset$new(
root = "/path/to/eicu-crd/2.0",
tables = c("diagnosis", "medication", "physicalexam"),
dev = TRUE
)
# Set mortality prediction task
task <- MortalityPredictionEICU$new()
sample_ds <- eicu_ds$set_task(task = task)
# View samples
head(sample_ds$samples)
} # }